Google Just Changed How It Sees Your Website — And Most SEOs Missed the Details That Matter

John Mueller’s latest Google Search News roundup dropped several updates that look routine on the surface. They are not.

Post Summary:

• Google Search Console now natively splits branded vs non-branded queries using AI — no more regex hacks
• Googlebot’s HTML fetch limit is confirmed at 2MB — silent truncation happens with no warning in GSC
• Vibe-coded (AI-built) websites are treated like any other site — but carry specific technical blind spots
• Google-Agent is a new crawler for AI agents hosted on Google infrastructure
• Universal Commerce Protocol (UCP) is live for e-commerce — an early signal of agentic shopping

I’ve been covering Google’s communication cadence for a while now, and there’s a pattern worth naming: the updates that change the most are announced in the most casual tone. Mueller broadcasts from Zurich, makes a joke, mentions something that sounds like housekeeping — and buried in the middle is something that changes how your site gets indexed or reported.

The latest Google Search News episode was exactly that. Here’s what actually matters, stripped of the PR wrapper.

Timeline: How These Updates Arrived

• November 2025 — Branded queries filter announced in Search Console; Query Groups launched
• February 2026 — Googlebot’s 2MB HTML fetch limit officially documented and clarified
• February 2026 — AI configuration tool for the Performance report rolled out globally
• March 11, 2026 — Branded queries filter expanded to all eligible sites
• March 31, 2026 — Google published detailed crawling documentation (“Inside Googlebot”)
• April 2026 — Google-Agent and UCP (Universal Commerce Protocol) announced for e-commerce

Search Console’s Branded Query Split: Finally Native

For years, separating branded from non-branded queries in Google Search Console required regex filters, custom Looker Studio dashboards, or third-party tools. Every approach was inconsistent. Every approach needed maintaining.

Google expanded its branded queries filter to all eligible sites on March 11, 2026, replacing the need for manual regex filters to separate brand traffic from organic discovery. The feature, originally announced in November 2025, uses an AI-assisted classification system that automatically identifies branded searches, including misspellings and brand-related product names.

What makes this genuinely different from a regex filter is the classification logic. A branded query is defined as one that includes your brand name, variations or misspellings of the brand name, and brand-related products or services. The classification is determined by an internal AI-assisted system — it includes your website brand name in all languages, typos, and also queries that don’t include the brand name but refer to a unique product or service of the site.

In practice: if your brand is Google, searches for “Gmail” now correctly register as branded — even though the word “Google” never appears in the query. A regex filter would have missed that entirely.

“Separating branded and non-branded queries has long required manual regex filters or keyword lists. This update gives you native segmentation in Search Console, making it easier to measure brand demand versus discovery traffic.” — Search Engine Land, March 2026

What this changes for reporting:

• Non-branded growth now has a clean, defensible baseline
• Brand awareness campaigns can be measured against branded search volume directly in GSC
• Branded queries typically lead to higher-ranking pages from your site and result in higher click-through rates, whereas non-branded queries offer organic growth, showing how new users find your content without any initial intent to go to your site

Practical tips:

• Check your Insights overview page first — there is now a card showing the branded/non-branded click split
• The filter only works on top-level domain properties, not subdomain or URL path properties
• Sites with insufficient query volume will not see the filter yet — consolidating to a domain-level property helps

Googlebot’s 2MB HTML Limit: The Silent Truncation Problem

This is the update most people dismissed as documentation housekeeping. It should not have been.

Googlebot currently fetches up to 2MB for any individual URL (excluding PDFs). For PDF files, the limit is 64MB. If your HTML file is larger than 2MB, Googlebot doesn’t reject the page — instead, it stops the fetch exactly at the 2MB cutoff. That downloaded portion is passed along to indexing systems as if it were the complete file. Any bytes that exist after that 2MB threshold are entirely ignored — they aren’t fetched, they aren’t rendered, and they aren’t indexed.

The reason this matters more than it sounds: Google Search Console reported no errors on oversized pages. Everything showed as “URL is on Google” and “Page is indexed” — everything looked perfectly normal. You get no warning. The page appears indexed. You have no idea the bottom half of your content does not exist as far as Google is concerned.

Testing confirmed this: A 3MB HTML file was truncated after 2MB — the source code cut off mid-word around line 15,210, literally stopping at “Prevention is b” and then the closing HTML tag.

Before you panic: According to the HTTP Archive Web Almanac 2025, the median HTML file size on mobile pages is a mere 33KB. Even at the 90th percentile, it’s only around 151KB. Only 0.82% of all analysed pages exceed 2MB.

The sites actually at risk are those with:

• Massive inline JavaScript or CSS embedded directly in HTML
• Large JSON objects loaded inline (common on e-commerce sites with product catalogs)
• Giant navigation menus in the HTML
• Inline base64 images

“Keep your HTML lean: Move heavy CSS and JavaScript to external files. Order matters: place your most critical elements — meta tags, title elements, canonicals, and essential structured data — higher up in the HTML document.” — Google Search Central Blog, March 2026

How to check your own site:

• In Chrome: open Developer Tools (F12) → Network tab → reload → check the HTML document size
• Use Screaming Frog, sorting by “Size” (uncompressed) descending — flag anything over 1MB for inspection
• Right-click any page → View Page Source → check file size

Vibe-Coded Websites: Fine for Search, With Caveats

“Vibe coding” — building websites primarily through AI tools with minimal manual code — has become mainstream enough that Google addressed it directly. Mueller’s position is clear and worth quoting precisely: these are essentially normal websites, and they can rank.

But the technical failure modes are consistent enough to be worth knowing. Mueller identified that one site stored key content in a llms.txt JavaScript file. He noted that Google doesn’t use this file, and he’s not aware of other search engines using it either. He recommended that the homepage should have everything that people and bots need to understand what the site is about.

The broader pattern across vibe-coded sites, as Mueller observed: “None of the flagged issues affects user experience. But every implementation choice Mueller criticised shares the same characteristic — it works for visitors while providing nothing to search engines.”

The specific checks Mueller recommends for AI-built sites:

• Verify rel="canonical" tags include the full URL including domain name
• Sites built on JavaScript frameworks (React, Next.js) require testing — verify Google can view the rendered page using the URL Inspection tool
• Add the site to Search Console immediately — before expecting traffic, not after
• Structured data is present on many vibe-coded sites already, which is genuinely useful

Google-Agent and UCP: The Agentic Commerce Signal

Two quieter announcements from the episode carry longer-term significance.

Google-Agent is a new crawler documented alongside existing bots for Read Aloud, NotebookLM, and Pinpoint. It is used by AI agents hosted on Google infrastructure. The introduction of Google-Agent marks a decisive turn towards agentic search and e-commerce. Site owners should prepare for AI agents like Google-Agent by reviewing structured data and ensuring sites are accessible to both users and AI crawlers.

Universal Commerce Protocol (UCP) creates a common language for AI agents interacting with e-commerce sites and businesses. Mueller’s framing in the episode was explicit: this is not something every e-commerce site needs to act on immediately. But the direction of travel is clear — shopping experiences are moving towards AI agent intermediation.

The practical implication now: structured data accuracy and product feed quality matter more as machine-readable signals, not less.

Comparing What Changed vs What Was Already True

What SEO Consultant Marie Haynes Has Observed

SEO professional Marie Haynes documented significant traffic gains across diverse website categories following the June 2025 core update, noting that successful content demonstrated comprehensive coverage anticipating follow-up questions and clear evidence of first-hand experience. This aligns directly with Mueller’s repeated framing in the episode: the question is not whether your site can be crawled — it is whether the content is worth crawling.

Mueller’s December 2025 Bluesky post makes this plainest. He endorsed an article describing most SEO content as “digital mulch” existing solely to “fill space, hit metrics, and appease the gods of Google.” His own guidance: write like blogging is alive, add value, and the technical infrastructure will support it — not replace it.

Frequently Asked Questions

Does the branded query filter affect how Google ranks my site? No. This filter is strictly a reporting feature and doesn’t affect how Google Search ranking works. It changes what you can see, not how Google evaluates your pages.

Will my site definitely be affected by the 2MB Googlebot limit? Almost certainly not. The median HTML page is around 30KB — you would need a page roughly 67 times larger than average to hit the threshold. But if you run a large e-commerce site with inline JSON product data or a content platform with bloated HTML, it is worth checking.

Does Google Search Console warn you when truncation happens? No. Testing has confirmed that GSC shows pages as “indexed” even when content has been silently cut off at the 2MB mark. The only way to verify is to inspect the page source size directly.

Do vibe-coded sites need special SEO treatment? No special treatment — but they carry specific failure patterns around JavaScript-dependent content, missing or incorrect canonical tags, and content stored in files Google does not read (like llms.txt). The standard checks apply with extra attention to those areas.

What should I do about UCP and Google-Agent right now? Ensure your structured data is accurate and your site is accessible without JavaScript where possible. UCP is early-stage — Mueller’s own advice was that not all e-commerce sites need to act immediately.

The Bottom Line

The pattern across all of these updates is the same one Mueller has been articulating for two years: Google is investing heavily in making its tools more readable for both machines and humans. The branded query filter makes human reporting cleaner. The crawl limit documentation makes machine behaviour transparent. The vibe-coding guidance makes the crawlability question answerable without guesswork.

None of it changes the fundamental calculation: content that adds genuine value, on a technically sound site, is what Search rewards. The tooling just makes it easier to see whether you are doing that or not.

Further reading:

• Google Search Console branded queries filter — official announcement
• Inside Googlebot: demystifying crawling and the 2MB limit
• Google’s SEO Starter Guide
• Google Search Central JavaScript SEO documentation